[opensuse-arm] Exclusive target for Raspberry PI
Hi: I was wondering.. it makes sense to create an exclusive target for the Raspberry PI (march=armv6zk -mcpu=arm1176jzf-s hard float, thumb) There are 1 millon devices out there..probably more than any other target we currently have .. is there any reason not to have it ? can anyone point me how exactly to integrate an armv6k compiler into the OBS ? Thanks and I am looking forward to your comments. -- To unsubscribe, e-mail: opensuse-arm+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-arm+owner@opensuse.org
On 28.12.2012, at 18:33, Cristian Rodríguez <crrodriguez@opensuse.org> wrote:
Hi:
I was wondering.. it makes sense to create an exclusive target for the Raspberry PI (march=armv6zk -mcpu=arm1176jzf-s hard float, thumb)
There are 1 millon devices out there..probably more than any other target we currently have .. is there any reason not to have it ? can anyone point me how exactly to integrate an armv6k compiler into the OBS ?
The main issue is our available build power. We are compiling everything on our existing x86 build farm, taking away precious time from our x86 builds. If we had native builds, that would allow us to have arm builds not conflict with non-arm. To do native builds however, we need virtualization to secure the build jobs. This is only possible as of Cortex A15. Machines with these are only barely coming out. We only have a single A15 board right now, which we have not enabled for native builds in obs yet. Alex -- To unsubscribe, e-mail: opensuse-arm+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-arm+owner@opensuse.org
2012/12/28 Alexander Graf <agraf@suse.de>: Hi Alex,
I was wondering.. it makes sense to create an exclusive target for the Raspberry PI (march=armv6zk -mcpu=arm1176jzf-s hard float, thumb) The main issue is our available build power. We are compiling everything on our existing x86 build farm, taking away precious time from our x86 builds.
Well, I'm not sure how many other ARMv5 devices we actually support with our ARMv5 tree. I think we decided originally for building ARMv5 due to things like sheevaplug and similar. Meanwhile however it is easy to say that the Raspberry PI is vastly more popular. So instead of adding an ARMv6 build, we could easily replace the ARMv5 build with an ARMv6 build. The question is then more how many "new" problems we will find and need to solve (miscompilations, configure scripts that can't deal with this particular target or the like).
From my unexperienced gut feeling I think it could be worth the effort.. all configure scripts I debugged had already special cases for armv6 inside, so overall I think it would be doable.
I guess the best way to test it is to build a toolchain for this triplet in e.g. Base:build:arm and then see from there. -- To unsubscribe, e-mail: opensuse-arm+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-arm+owner@opensuse.org
Am Mittwoch, 2. Januar 2013, 13:14:14 schrieb Dirk Müller:
2012/12/28 Alexander Graf <agraf@suse.de>:
Hi Alex,
I was wondering.. it makes sense to create an exclusive target for the Raspberry PI (march=armv6zk -mcpu=arm1176jzf-s hard float, thumb) The main issue is our available build power. We are compiling everything on our existing x86 build farm, taking away precious time from our x86 builds.
Well, I'm not sure how many other ARMv5 devices we actually support with our ARMv5 tree. I think we decided originally for building ARMv5 due to things like sheevaplug and similar. Meanwhile however it is easy to say that the Raspberry PI is vastly more popular.
So instead of adding an ARMv6 build, we could easily replace the ARMv5 build with an ARMv6 build. The question is then more how many "new" problems we will find and need to solve (miscompilations, configure scripts that can't deal with this particular target or the like).
From my unexperienced gut feeling I think it could be worth the effort.. all configure scripts I debugged had already special cases for armv6 inside, so overall I think it would be doable.
I guess the best way to test it is to build a toolchain for this triplet in e.g. Base:build:arm and then see from there.
We could also disable armv5 build to save resources and start an armv6 one in openSUSE:Factory:ARM. Since we still have some more days without high load, I am all for trying this starting today. Any concerns? -- Adrian Schroeter SUSE Linux Products GmbH email: adrian@suse.de -- To unsubscribe, e-mail: opensuse-arm+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-arm+owner@opensuse.org
Am Mittwoch, 2. Januar 2013, 13:28:33 schrieb Adrian Schröter:
Am Mittwoch, 2. Januar 2013, 13:14:14 schrieb Dirk Müller:
2012/12/28 Alexander Graf <agraf@suse.de>:
Hi Alex,
I was wondering.. it makes sense to create an exclusive target for the Raspberry PI (march=armv6zk -mcpu=arm1176jzf-s hard float, thumb) The main issue is our available build power. We are compiling everything on our existing x86 build farm, taking away precious time from our x86 builds.
Well, I'm not sure how many other ARMv5 devices we actually support with our ARMv5 tree. I think we decided originally for building ARMv5 due to things like sheevaplug and similar. Meanwhile however it is easy to say that the Raspberry PI is vastly more popular.
So instead of adding an ARMv6 build, we could easily replace the ARMv5 build with an ARMv6 build. The question is then more how many "new" problems we will find and need to solve (miscompilations, configure scripts that can't deal with this particular target or the like).
From my unexperienced gut feeling I think it could be worth the effort.. all configure scripts I debugged had already special cases for armv6 inside, so overall I think it would be doable.
I guess the best way to test it is to build a toolchain for this triplet in e.g. Base:build:arm and then see from there.
We could also disable armv5 build to save resources and start an armv6 one in openSUSE:Factory:ARM.
Since we still have some more days without high load, I am all for trying this starting today.
Any concerns?
Hm, speaking about this here over a coffee, we are unsure which improvements we actually would get by switching to armv6 (which can not be achieve by runtime optimizations). But we would definitive loose hardware support by dropping armv5. So we like to stay on armv5 for now. But everybody should be able to rebuild our armv5 distro with armv6 flags in an own OBS instance. So just try that and make some measurements :) -- Adrian Schroeter SUSE Linux Products GmbH email: adrian@suse.de -- To unsubscribe, e-mail: opensuse-arm+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-arm+owner@opensuse.org
El 02/01/13 09:14, Dirk Müller escribió:
So instead of adding an ARMv6 build, we could easily replace the ARMv5 build with an ARMv6 build. The question is then more how many "new" problems we will find and need to solve (miscompilations, configure scripts that can't deal with this particular target or the like).
I will check the fallout ;)
all configure scripts I debugged had already special cases for armv6 inside, so overall I think it would be doable.
Well, special (broken) cases are found everywhere.. usually to workaround different kinds of brokeness. THose are just poorly written.. -- To unsubscribe, e-mail: opensuse-arm+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-arm+owner@opensuse.org
participants (4)
-
Adrian Schröter -
Alexander Graf -
Cristian Rodríguez -
Dirk Müller